Post Your NEW 3DMark Sky Driver Benchmark Results

I don't personally agree with that but at the same time I think its been a very long time if ever 3dmark has ever graded the cpu power and gpu power it seems equally...Some would argue hyper threading shouldn't be relevant in coming up with a physics score...everyone is entitled to there opinion...I have turned it off to play certain games and just found it to get less performance
It seems the people getting very high scores have very powerful cpus and gpus where as some of the older 3dmark test really only tested the gpus.....is pretty much my take on it

I kind of understand the CPU scoring part of it. It's the GPU part that's throwing me off. like one 290x is right there with my gtx780 sli (cpu score similar between the two) I guess those few extra cpu points really add to your score. I understand that 290x>780 , but normally wouldn't 2x 780 > 1x 290x? Or is my CPU really limited my SLI performance that much?

It's a different story when I run firestrike. The numbers all seem to line up then. That's why I think this particular bench is just wonky with high end rigs. Just my opinion. :)
 
http://www.3dmark.com/sd/2186254

3770K@ 4.5Ghz, 16GB RAM 1600mhz, 660TI FTW Stock core 1315mhz Memory 7008mhz

1z6ruic.jpg



Also:
Fire Strike

Cloud Gate

Ice Storm
 
Last edited:
I kind of understand the CPU scoring part of it. It's the GPU part that's throwing me off. like one 290x is right there with my gtx780 sli (cpu score similar between the two) I guess those few extra cpu points really add to your score. I understand that 290x>780 , but normally wouldn't 2x 780 > 1x 290x? Or is my CPU really limited my SLI performance that much?

It's a different story when I run firestrike. The numbers all seem to line up then. That's why I think this particular bench is just wonky with high end rigs. Just my opinion. :)

the bottom line is this...your much newer cpu scores around 30% lower than mine...but not in just this test...probably any test that uses multi cores/ threads...skydiver sees this as a bottleneck and drops your score accordingly...im personally glad they finally come out with a test that pushes both...and why is that? because in my opinion a real game should push both......one of my favorite games is bf4...but it only uses like 30% of my cpu while pushing my gpu at 100%...thats telling me im gpu bottle necked.... but i feel like the game could do so much more if it used the cpu to the fullest as well...possibly if i had a 780ti or 290x it would push the cpu harder....its possible i guess

firestrike is a gpu test only...i can see someone scoring a super high score and still not being able to play certain games cause there cpu is to weak
 
the bottom line is this...your much newer cpu scores around 30% lower than mine...but not in just this test...probably any test that uses multi cores/ threads...skydiver sees this as a bottleneck and drops your score accordingly...im personally glad they finally come out with a test that pushes both...and why is that? because in my opinion a real game should push both......one of my favorite games is bf4...but it only uses like 30% of my cpu while pushing my gpu at 100%...thats telling me im gpu bottle necked.... but i feel like the game could do so much more if it used the cpu to the fullest as well...possibly if i had a 780ti or 290x it would push the cpu harder....its possible i guess

firestrike is a gpu test only...i can see someone scoring a super high score and still not being able to play certain games cause there cpu is to weak

My cpu is only a year newer then yours :D Intel shows yours was Q1-2010 and mine was Q1-2011. I wish I would have waited and picked up an x79 board/cpu. I'm waiting to see how the new 8 core Intel's look....then maybe.....

You should rerun the test now to see if your score increases any. Mine went up a good amount for some reason.
 
the bottom line is this...your much newer cpu scores around 30% lower than mine...but not in just this test...probably any test that uses multi cores/ threads...skydiver sees this as a bottleneck and drops your score accordingly...im personally glad they finally come out with a test that pushes both...and why is that? because in my opinion a real game should push both......one of my favorite games is bf4...but it only uses like 30% of my cpu while pushing my gpu at 100%...thats telling me im gpu bottle necked.... but i feel like the game could do so much more if it used the cpu to the fullest as well...possibly if i had a 780ti or 290x it would push the cpu harder....its possible i guess

firestrike is a gpu test only...i can see someone scoring a super high score and still not being able to play certain games cause there cpu is to weak

BF4 does a pretty good job of scaling with cores and in my oppinion----doesn't show it's best colors until you have 6 and 8 core CPUs running it. It may not slam your CPU 100%----but that's because you are spreading it out over 6 threads. It's a more efficient setup and also should net more conistant and/or lower frametimes. Which is important for smoothness and responsiveness.

Go jump on a 24/7 siege of shanghai server and bench FRAPS for framerate and frametimes with 4 cores/threads and then 6 cores/threads. 4 cores will have higher CPU utilization, but 6 cores will perform better. If you can't disable cores in your bios, then you can at least set core affinity in windows task manager.

Rather interesting is that from what I've seen, BF4 doesn't perform as well on hyperthreading, if you have at least 4 cores to begin with.
 
Last edited:
https://www.dropbox.com/sh/wm37etee63j12oe/AACGD8lCkPvn-BFY2Gf7Vum3a

that's a 2 image slideshow of my results. The one that says "valid" is AMD's 14.4 whql driver. The one that says "not approved" is 14.6-RC2.

That system is the following:

HD7870 - 1150 core/1325memory
AMD Phenom II X6 - 3.7GHz
8GB DDR3 - 1772GHz

Motherboard Northbridge is 2659GHz, which is up from 2000GHz (overclocking the northbridge helps Phenom II performance a ton).
 
Yeah, I was noticing that. Comparing our physics scores (and only our physics scores, since they're totally CPU dependent):

Core i5 2500k @ 4.5 GHz = 8244 points
Core i7 4770k @ 4.5 GHz = 11973 points

That's a 45% increase in performance at the same clockspeed... which is pretty nuts. The only thing I can think of that would put such a massive gap between these two chips is if 3DMark seriously loves HyperThreading.

Was hyperthreading running on that 4770k during that run?

Some games benefits from hyperthreading, some do not. A 4770K is a more refined design than the 2500k and has more cache. It's easily possible the performance increase is from the fact it's an i7, a newer design, has more cache. More emphasis on the newer design as 4 core Haswell i7 even beats a 4 core Ivy Bridge E even though the Ivy Bridge E has more cache.

I'd like to see hyperthreading VS. no hyperthreading, with this benchmark.
 
Was hyperthreading running on that 4770k during that run?

Some games benefits from hyperthreading, some do not. A 4770K is a more refined design than the 2500k and has more cache. It's easily possible the performance increase is from the fact it's an i7, a newer design, has more cache. More emphasis on the newer design as 4 core Haswell i7 even beats a 4 core Ivy Bridge E even though the Ivy Bridge E has more cache.

I'd like to see hyperthreading VS. no hyperthreading, with this benchmark.

i've tested for you. :cool:

this its my first result, With Hyper threading ON 3770k 4.5ghz, 660TI FTW stock clock(max boost clock of 1320mhz) memory 7008mhz.

http://www.3dmark.com/sd/2186254

Overall: 20527
Graphics: 23711
CPU: 11764
Combined: 23082


This its my second result, With Hyper Threading OFF 3770k 4.5ghz, 660TI FTW stock clock(max boost clock of 1320mhz) memory 7008mhz.

http://www.3dmark.com/3dm/3433529

Overall: 19062
Graphics: 23899
CPU: 9275
Combined: 20393
 
So basically, this benchmark really loves Hyperthreading :p

That's a 26% increase in CPU score just by turning on HT between those two runs :eek:
 
that is what its supposedly to do Hyper Threading. between 20% to 30% of increased performance depending of the application by just using up to 5% more resources..
 
Thanks Araxie! That's a very informative comparison.

Now I wonder if someone can compare a 6 core Intel with and without hyperthreading??? Curious to see how many threads this thing scales with.
 
You guys know that newer nvidia drivers are messed up when running Sky Diver right? I get a black screen on one of the graphics tests and my bench score is shit. Like SLI is not working or something.

I think it may be an issue with 3Dmark itself. I got a black screen one time after I switched my AMD HD7870 from 14.4 whql to 14.6 RC2 betas. and, there are certain parts which aren't very smooth, every single time I run the benchmark. I think I may go back to the 14.4 whql...
 
Hmm maybe its my 6 Core CPU helping me out looks like this banchmark loves more cores.

of course.. with that difference of chips you need to compare its Graphic Score and not the Overall score unless you have a similar processor.. I with a 660TI have better overall Score better than people here with i5 + gtx 770 just for the fact that the the overall score its heavily affected by the chip threads..
 
that is what its supposedly to do Hyper Threading. between 20% to 30% of increased performance depending of the application by just using up to 5% more resources..
Sure, though games generally don't see any gains from HT at all.

Bit weird to see a gaming benchmark weigh it so heavily.
 
Sure, though games generally don't see any gains from HT at all.

Bit weird to see a gaming benchmark weigh it so heavily.

I think the real gaming its just because its another level really, its a bit unfair (I think) make games like crysis 3 for example.. where a 4 cores i7 crush any i5 by a large amount, but just not that it also limit and bottleneck the GPU... so, want developers really to all the games perform like that? many people are still even using Dual Core chips.. and of course i3 chips which are heavily affected by the Hyperthreading.. I guess with the time we will see more games making a good use of HT as Crysis 3 do, but that will be very slowly, i guess thinking as developer its hard to push people to buy i7 chips to enjoy the fully potential of a game.. how much people can every year spend ~500$ in GPU + ~300CPU + **insert a good mobo average 150$ here** and thats forgetting high end 6 Core Intel Chips + High end motherboard + high end Xfire or SLI setups + high end PSU and all the cooling involved to maintain that setup its a lot of money.. if developers push that hard PC gamers would happen the same as the first Crysis in that time that game was only to enjoy by few people for some people was a dream to play crysis decently, others were proud to play crysis and say to their friends "see my uber high end machine playing smooth crysis maxed".. and for others was like a benchmark..
 
http://www.3dmark.com/3dm/3439637?

Graphics 61138
Physics 13927
Combined 23897

3930k @ 4.2 and 290's @1050 ... I think I have a bit left. Test wasn't maxing the video cards out at all.. When I get my water cooling setup back up and my 3930k back at 4.7ghz I think I'll be able to squease out a few more points.
 
Bumped the gpu overclock a tad more and disabled all power saving in the bios has increased my score a small amount...kinda proud my old intel still owne's the newest amd heavily overclocked 8 core cpus...it is what it is;)
Capture_zps07d195e1.jpg
[/URL][/IMG]
 
Wow, seems like this benchmark is VERY well threaded!
Or it's handing out an arbitrary ~25% bonus because the reported number of cores is higher.

Someone should run FRAPS with HT off and HT on, see if the bench actually runs ~25% faster.
 
Or it's handing out an arbitrary ~25% bonus because the reported number of cores is higher.

Someone should run FRAPS with HT off and HT on, see if the bench actually runs ~25% faster.

i can do that if you wish...a bit latter on
 
Ok i have to ask.. why are some 4770K scoring so low?.. Unstable overclock or what?.. are people with 4770K disabling HyperThreading or what?.. I score with my 3770K in physic test [email protected]: 11764 and combined Score: 23082 but i see here people with [email protected] scoring even less than 11500 and a couple with less than 11000 and horrible Combined results even with multiple GPU Setup.. but i see others [email protected] scoring 11500+ and average Combined results.. its really a thing of unstable overclock?..
 
Or it's handing out an arbitrary ~25% bonus because the reported number of cores is higher.

Someone should run FRAPS with HT off and HT on, see if the bench actually runs ~25% faster.

its really working faster.. and its actually using at a higher percent the cores.. i have in my G510 some monitoring apps, Aida64, Core temp, and some monitoring real time with my android phone at the moment of the test.. so yes, its actually using fully the 8 threads at a higher percent.. 90-99%.. and yes the GPU work better and faster i use in a second screen the EVGA precision X hardware monitor and are some serious differences between both at 4c/8t and 4c/4t..
 
Or it's handing out an arbitrary ~25% bonus because the reported number of cores is higher.

Someone should run FRAPS with HT off and HT on, see if the bench actually runs ~25% faster.
I don't think that it is, as my 6 core Phenom II get's what it deserves.


No need for fraps, even the free version of this test gives you framerate numbers in the online page it directs you to.
 
Back
Top